Meta Ads Healthcheck

Required Credentials

Credential	Where to Get	Used For	OAuth Scope
META_ACCESS_TOKEN	Meta Developer Console → Graph API Explorer → Generate Token	Fetching campaign and adset data	ads_read (read-only)
META_AD_ACCOUNT_ID	Ads Manager URL: adsmanager.facebook.com/act_XXXXXXXXX	Identifying which account to query	—

When to Use

When user wants a quick "are my ads working?" check — not a deep diagnostic, just a status update. Use before a meeting, or any time something feels off.

	meta-ads-healthcheck	meta-ads-daily-pulse
Primary question	"Is this campaign healthy or not?"	"Did anything change vs. last week?"
Comparison basis	Fixed Green/Yellow/Red thresholds	Same day of prior week (WoW)
Best for	On-demand check when something feels wrong; meeting prep	Daily routine, morning scan
Output style	Campaign-by-campaign status report	Change alerts ranked by revenue impact

---

Step 1: Get the Basics

Pull the user's Meta Ads data for today and the past 7 days:

• Campaign name, status, spend
• Results, cost per result
• CTR, frequency
• Any alerts or issues (delivery low, learning limited, paused)

Use Meta Marketing API to fetch campaign and adset level data.

---

Step 2: Traffic Light Assessment

Score each campaign as Green / Yellow / Red:

Signal	Green	Yellow	Red
Delivery	Spending normally	< 80% of budget	< 50% or zero
Cost per result	At or below target	10–30% above target (and variance is normal)	30%+ above target and sustained (not just a spike)
Frequency	< 3	3–4	> 4
Learning	Out of learning	In learning (< 7 days)	Stuck in learning
CTR	> 1%	0.5–1%	< 0.5%

---

Step 3: Check for Normal Performance Fluctuations

Before flagging a campaign as Yellow or Red, confirm it's not normal fluctuation.

Normal fluctuation — monitor only:

• CPA is bouncing around day-to-day (up to ±30%) but the rolling average is stable — this is normal Meta Ads volatility, not a problem
• Weekend vs. weekday differences
• Gradual changes over weeks
• Variation while in Learning Phase

Concerning — investigate today:

• CPA average has shifted up and stayed there for 3+ consecutive days (not just a one-day spike) — this is a real trend, not fluctuation
• Delivery dropping to near zero with no budget change
• Conversion rate declining while spend increases
• Performance degradation after a recent edit

Before diagnosing problems, ask:

1. Is the ad set still in Learning Phase? (if yes, delay judgment — data is unstable)
2. What's the baseline for normal variation for this campaign?
3. Are there external factors — seasonality, competitor activity, or platform changes?
4. Is sample size sufficient? (typically need 7+ days of data for stable ad sets)

---

Step 4: Surface Key Issues

For each Yellow or Red campaign, identify:

• What is the specific problem?
• How urgent is it? (Needs attention today / this week / monitor)
• What likely caused it?

Creative Refresh Trigger — Frequency > 3: When frequency reaches 3 or higher, flag the campaign as "Creative Refresh Needed": "Campaign [X] has frequency at [N] — audience is seeing the same ads too often. This typically causes CTR to drop and CPL to rise. Creative refresh is indicated."

Note: When this flag appears, the appropriate next step is meta-ads-recommendation — it will prescribe the specific creative refresh action based on the full analysis context. Do not trigger meta-ads-builder directly from this skill.

---

Step 5: Quick Output

Overall Status: [N] campaigns running | [N] Green | [N] Yellow | [N] Red

For each Red/Yellow campaign:

• Campaign name
• Status: [Issue]
• Spend / Results / CPL
• What happened (1 sentence)
• Suggested focus area (1 sentence)

Everything else: brief healthy summary.

---

Priority Sort — What to Check First

When you have 10+ active campaigns:

1. Red campaigns with spend > $50/day → Check today
2. Yellow campaigns where CPL is 20%+ above target → Check today
3. New campaigns (< 7 days, still in Learning) → Flag as "Learning Phase — Do Not Judge Yet"
4. Green campaigns → Review weekly

---

When to Escalate

• If the health check surfaces a real problem: suggest meta-ads-analysis for a full diagnostic
• For daily routine monitoring: suggest meta-ads-daily-pulse instead

---

Skill Boundary

• This is a quick scan, not a deep diagnostic — don't draw conclusions beyond what's clearly broken or healthy
• Do not make optimization recommendations in this skill — that's meta-ads-recommendation's job
• Do not re-fetch data if it was recently pulled in the same session